**Assessment of the Quality of Polluted Areas in Northwest Romania Based on the Content of Elements in Di**ff**erent Organs of Grapevine (***Vitis vinifera* **L.)**

#### **Florin Dumitru Bora 1, Claudiu Ioan Bunea 2, Romeo Chira <sup>3</sup> and Andrea Bunea 4,\***


Received: 19 January 2020; Accepted: 6 February 2020; Published: 9 February 2020

**Abstract:** The purpose of this study was to evaluate the environmental quality of polluted areas near the Baia Mare Mining and Smelting Complex for future improvements the quality of the environment in polluted areas, such as the city of Baia Mare and its surroundings. Samples of soil and organs of grapevine (*Vitis vinifera* L.) were collected from Baia Mare, Baia Sprie and surrounding areas (Simleul Silvaniei) and their content of Cu, Zn, Pb, Cd, Ni, Co, As, Cr, Hg were analyzed. Most soil and plant samples showed higher metal concentrations in Baia Mare and Baia Sprie areas compared to Simleul Silvaniei, exceeding the normal values. The results obtained from the translocation factors, mobility ratio, as well as from Pearson correlation study confirmed that very useful information is recorded in plant organs: root, canes, leaves and fruit. Results also indicated that *Vitis vinifera* L. has some highly effective strategies to tolerate heavy metal-induced stress, may also be useful as a vegetation protection barrier from considerable atmospheric pollution. At the same time, berries are safe for consumption to a large degree, which is a great advantage of this species.

**Keywords:** heavy metals; grapevine; bioaccumulation; biomonitoring

#### **1. Introduction**

Pollution is a worldwide problem caused by anthropogenic activities such as mining, petrochemical refining, and smelting, with negative impacts on human health. In Romania, 18% of population was exposed to heavy environmental pollution whereby serious health risks are likely. A total of 14 environmental pollution "hot spots" have been identified in Romania: Cops, a Mică, Baia Mare, Ploies,ti-Brazi, Zlatna, Ones,ti, Bacău, Suceava, Petes,ti, Târgu Mures, Turnu Măgurele, Talcea, Isalnita, Bras, ov, and Govora; 5.3% of the population lives in these heavily polluted areas, mostly in the critical rural/urban interfaces [1–3].

Growing in extremely polluted areas, some plant species can be seriously damaged, whereas others can survive without any visible changes [4]. Uptake of trace metals by plants can happen from the soil through the roots and subsequent transport to the leaves or directly from the air. Specific mechanisms allow plant tissues to accumulate high quantities of trace metals, playing, thereby, a vital role in the natural recovery of industrial damage [5]. In this respect, trees are especially useful because contaminants can accumulate in their large biomass and they can grow in soil with poor fertility and

structure [6]. Terrestrial higher plants are specific living-system structures with unique ecobiological characteristics. They interact actively with three spheres: soil, water and air, at the same time, requiring only modest nutrient input. Along with nutrients, plant roots can absorb a range of anthropogenic toxic materials. Heavy metals are just a class of such pollutants and several of them are well known as nonessential and extremely toxic for plants: cadmium (Cd), lead (Pb), mercury (Hg) and arsenic (As). Even essential micronutrients such as copper (Cu), zinc (Zn), and nickel (Ni) may become toxic for plants when absorbed above certain threshold values [7].

Plants have developed effective detoxification mechanisms to manage heavy metal content [8]. Some species may concentrate heavy metals in root cell walls and/or vacuoles, thus minimizing their phytotoxicity [9] and also preventing the spread of these contaminants in soil [10]. Phytoremediation is an excellent opportunity for cleaning up the pollute environment in an economic and ecological friendly manner. It uses green plants to detoxify the polluted environment, and it may be applied in a variety of ways [10]. On the other hand, plants can be used as indicators of the pollution level of the environment. Heavy metals in plant organs, especially in roots and leaves, represent a very specific evidence of spatial and temporal history of polluted area [11]. Researchers agree that the root and leaf analyses are essential in the evaluation process of the environmental quality of ecosystems or to study the effects of heavy metals on the chemical composition of plants.

Grapevine is an important crop worldwide, while the wine sector is of major importance for the economy of many countries [12]. The soil chemistry in vineyards influences wine and grape quality, vine-soil relationship being a key part of the concept of terroir [13,14]. The town of Baia Mare has been an important nonferrous metallurgical center where heavy metals like Pb and Cu have been extracted and processed for centuries. Metallurgical plants "Romplumb", located in the Ferneziu district, and "Cuprom", located in the eastern part of the city, had polluted the soil in Baia Mare area with Pb, Cd, Cu, Zn, and As [15–17].

In this study, concentration of Cu, Zn, Pb, Cd, Ni, Co, As, Cr and Hg in vineyard soil, several parts of grapevine (*Vitis vinifera* L.), as well as in must and wine from Baia Mare, Baia Sprie, and Simleul Silvaniei areas were analyzed.

#### **2. Results and Discussion**

#### *2.1. Metal Concentration in Soil Samples*

Elemental concentration varied among soil samples but were considerably higher than concentrations allowed by the Romanian Regulation of allowable quantities of hazardous and harmful substance in soil (Order of the Ministry of Waters, Forests and Environmental Protection No. 756/3 November 1997), as well as by the Council Directive 86/278/EEC for Protection of the Environment (European Communities Council 1986) (Table 1). Physical properties of soil samples are provided in Supplementary Table S1.

Regardless of sampling depth, the highest concentrations of Cu were recorded in Baia Sprie area, followed by the Baia Mare area and Simleul Silvaniei area (Table 1). In all cases, the concentrations significantly exceeded the normal values set by the corresponding legislation (20 mg/kg). These high concentrations can be attributed to the pollution factor (in Baia Mare and Baia Sprie areas) or the extensive usage of Cu-based plant protection products (in the Simleul Silvaniei area). Detected values were higher than those reported previously from this area (640.6 mg/kg) [17], (599.75 mg/kg) [18], (314.00 mg/kg) [16] or other wine-producing areas in Southeast Romania [19], but were within the range established for Cops, a Mică (77–7675 mg/kg) [3].


Thecontentofheavymetalsinsoilfromareasstudied(mg/kgDW)(Mean±standarddeviation)


**Table 1.** *Cont*. difference (*<sup>p</sup>* ≤ 0.05) between the depths of the soil profile. The difference between any two values, followed by at least one common letter, is insignificant. \*Order of the MinistryWaters, Forests and Environmental Protection No. 756/3 November 1997, approving the regulation on the assessment of environmental pollution, Bucharest, Romania; 1997. \*\*M.A.L(Maximum Admissible Limit) = Normal Values. in = insignificant

The results obtained by Damian et al. (44–5823 mg/kg) are comparable to those obtained in this research. The Cu values obtained for Simleul Silvaniei are conformable with those recorded by Alagi´c et al. (293.00 mg/kg) [21] and Bora et al. (479.64 mg/kg) [23].

In the Baia Sprie and Baia Mare areas, the concentrations of Zn tended to increase with the sampling depth, with the highest concentrations being detected in samples collected at 60–80 cm (3483.25 ± 94.11 mg/kg and 2734.93 ± 147.45 mg/kg, respectively). All values greatly exceeded the normal levels of Zn allowed by the law (100 mg/kg). In contrast, in the Simleul Silvaniei area, the highest concentration was recorded in the surface soil profile (76.86 ± 7.71 mg/kg (0–20 m)), and it tended to decrease with the increasing soil depth. The Zn values obtained are higher than data published in previous reports from Baia Mare or other Romanian regions [16,26,27].

Concentrations of Pb and Cd varied within a wide range (Table 1). The highest concentrations were recorded in the Baia Mare and Baia Sprie areas, significantly higher than those detected in Simleul Silvaniei or allowed by the applicable legislation (20 mg/kg for Pb, 1 mg/kg for Cd). The extremely high values of Pb and Cd indicate severe heavy metal pollution in these two areas. Similar [18] or lower [16,27] values were also reported from these regions. The average content of Ni and Co in soil samples exceeded the normal concentration in the Baia Mare area (25.29 ± 2.07 vs 20 mg/kg for Ni and 22.57 ± 1.65 vs 15 mg/kg for Co), but were below the limit in the other two areas (Table 1). For comparison, Mihali et al. recorded similar values (13.1 mg/kg [Ni] and 24.8 mg/kg [Co]) in the Baia Mare area [19], while another study conducted in unpolluted regions from Dobrogea and Muntenia reported Ni concentrations between 0.97–11.29 mg/kg and Co concentrations between 0.49–4.36 mg/kg [26].

The concentrations of As, Cr, and Hg indicated no pollution of the soil samples with these heavymetals; values were below the normal levels. The highest values were obtained in Baia Sprie (4.46 ± 1.56 mg/kg As; 2.31 ± 0.78 mg/kg Cr; 0.064 ± 0.016 mg/kg Hg) followed by Baia Mare (3.46 ± 0.63 mg/kg As; 2.51 ± 0.51 mg/kg Cr; 0.052 ± 0.021 mg/kg Hg) area. A recent study conducted in Vaslui county reported higher content of As (10.14 mg/kg) and Cr (62.05 mg/kg) compared to our results [28].

#### *2.2. Metal Concentration in Plant Material Samples*

#### 2.2.1. Metal Concentration in Roots

Roots are in direct contact with the soil solution and the concentration of heavy metals in roots is generally used as indicative of soil metal bioavailability [29]. Varieties cultivated in Simleul Silvaniei showed the lowest concentrations of Cu and Zn, compared to the varieties from Baia Mare and Baia Sprie. Italian Riesling from Baia Mare and Baia Sprie (779.15 ± 4.66 mg/kg and 670.51 ± 6.61 mg/kg, respectively) and Feteasca alba from Baia Sprie (669.15 ± 21.27 mg/kg) contained the highest concentration of Cu while the varieties cultivated in Baia Mare area recorded the highest Zn concentration (Table 2). Studies have shown that high concentration of Cu can affect the growth of the roots [9,30]. The highest concentrations of Pb was registered in Feteasca regala from Baia Mare (60.81 ± 5.95 mg/kg), in Italian Riesling from Baia Sprie (92.26 ± 1.11 mg/kg), significantly higher as compared to the same varieties grown in the Simleul Silvaniei area (0.83 ± 0.60 mg/kg (Feteasca regala) and 0.43 ± 0.17 mg/kg (Italian Riesling)). According to Vamerali et al., Pb has no important role in functions of plants [7]. Roots of Feteasca regala from Baia Mare and Feteasca alba from Baia Sprie had the highest concentration of Cd (7.09 ± 0.83 mg/kg and 3.07 ± 0.12 mg/kg, respectively) and Co (32.24 ± 1.23 mg/kg and 10.95 ± 1.26 mg/kg) (Table 2). Cd and Co concentrations in these areas were significantly higher than those obtained in varieties grown from Simleul Silvaniei. In Baia Sprie and Simleul Silvaniei, concentrations of Ni and As were similar amongst varieties, while in Baia Mare area, Italian Riesling variety had higher concentration of As compared to Feteasca alba and Feteasca regala (Table 2). Interestingly, roots from Simleul Silvaniei showed higher content of Ni compared to varieties from Baia Sprie. Concentration of Cr was similar for all three varieties in Baia Mare and Baia Sprie, except for Feteasca regala from Baia Mare which recorded a significantly higher concentration.



#### *Molecules* **2020**, *25*, 750


**Table 2.** *Cont*.


difference (*<sup>p</sup>* ≤ 0.05) between the plant parts of the same variety. The difference between any two values, followed by at least one common letter, is insignificant.aVameralial.,2010[7];bAlloway,2013[31];cKabata-Pendias,2010[25].

 et

#### *Molecules* **2020**, *25*, 750

**Table 2.** *Cont*.

The varieties grown in Simleul Silvaniei have a lower concentration compared to varieties grown in Baia Mare and Baia Sprie. Hg was detected in low concentrations in all three areas. The observed concentrations of Cu, Zn, Ni, and As exceed the toxic threshold in plant tissues [7,31]. Overall, data suggests that high concentrations of heavy metals in soil result an increased metal content in the roots as well. Compared to our findings, grapevines grown in polluted areas from East Serbia have shown similar concentrations of heavy metals in roots [21].

#### 2.2.2. Metal Concentration in Canes

Cu has the highest concentrations in all varieties cultivated in Baia Sprie (147.92 ± 2.46 mg/kg (Italian Riesling); 143.72 ± 2.46 mg/kg (Feteasca Regala) 124.56 ± 9.02 mg/kg (Feteasca Alba), followed by the Baia Mare area (77.31 ± 3.76 mg/kg (Feteasca Regala); 72.93 ± 2.15 mg/kg (Italian Riesling); 72.03 ± 2.75 (Feteasca Alba). This can be explained with the heavy metal pollution phenomenon. Though Cu is involved in many vital processes in plants such as photosynthesis, flowering, seed production, and plant growth, its excessive concentrations may cause a significant modification of biochemical processes, leading to the reduction of shoot growth [7,32]. Results obtained in Baia Mare are comparable with those reported from Turulung, NW Romania (63.67 ± 2.67 mg/kg) [32] and much lower than those obtained from polluted regions from East Serbia (170.90 ± 0.80 mg/kg Cu [Flotacijsko Jalovište]; 175.00 ± 2.00 mg/kg [Bolniˇcko naselje]; 160.00 ± 0.90 mg/kg [Slatinsko naselje]) [21]. Analyzing the concentration of Zn in the canes, varieties cultivated in Baia Sprie had the highest concentrations. Cane samples from the Baia Mare area also displayed high concentrations of Zn, exceeding the toxicity threshold in plant tissues [7,31]. The lower concentrations detected in the Simleul Silvaniei area were consistent with the literature values reported for other areas [21,23]. Regarding the concentration of Pb, Cd, Ni, and Co in the string and canes, values recorded in the Baia Sprie and Baia Mare areas are significantly higher than those recorded for the same heavy metals in the Simleul Silvaniei area (Table 2) or reported from other regions [21,23]. In all regions and varieties studied, concentrations of As, Cr and Hg were similar and below the toxicity threshold in plant tissues.

#### 2.2.3. Metal Concentration in Leaves

Agricultural crops are especially sensitive to Cu concentration. As a first signal of excessive supply of Cu, symptoms of chlorosis may occur [32]. In this study, significantly higher concentrations of Cu were detected in Baia Sprie as compared to Baia Mare. No significant differences in Cu concentrations were found between varieties cultivated in the same area, except for Simleul Silvaniei, where leaves of Feteasca regala had higher Cu content as the other two varieties tested. Similarly, Zn concentrations were highest in Baia Sprie. Feteasca alba leaves from Baia Sprie and Italian Riesling leaves from Baia Mare had significantly higher content of Zn than leaves of other varieties collected from the same area. For both Cu and Zn, concentrations were above the phytotoxic threshold. Concentrations of Pb, Cd, and Ni in leaves were similar across varieties cultivated in the same area. While Cd and Ni did not exceed the phytotoxic concentrations established for plant tissues, the concentrations of Pb in leaves collected from Baia Sprie and Baia Mare areas were greatly above the pre-defined phytotoxic concentration, which can be attributed to the pollution factor in these areas. Concentrations of Co were similar in Baia Mare and Baia Sprie areas ranging between 4.84 ± 0.88 mg/kg and 6.89 ± 1.01 mg/kg (Table 2). The levels in leaves were slightly above the normal range [25], but still below the phytotoxic concentration.

#### 2.2.4. Metal Concentration in Grapes

According to Vamerali et al., Cu is a constituent of enzymes involved in photosynthesis, in reproductive phase, and in determining the yield and quality in crops. Zn is a constituent of cell membranes and it is involved in DNA transcription, activation of enzymes, and evaluation of the yield and quality of crops [7]. Varieties cultivated in Baia Mare and Baia Sprie areas recorded comparable concentrations of Cu (8.49 ± 0.64–12.30 ± 2.39 mg/kg and 12.31 ± 1.82–14.51 ± 1.25 mg/kg, respectively) and Zn (8.52 ± 1.25–10.13 ± 1.33 mg/kg and 6.78 ± 2.14–8.91 ± 1.50 mg/kg), but values were comparable

and within the normal range accepted in plant tissues (Table 2). While these concentrations can be attributed to the heavy metal pollution phenomenon in the two areas, Cu and Zn content of varieties cultivated in Simleul Silvaniei area (2.31 ± 0.77–3.38 ± 1.76 mg/kg and 1.11 ± 0.62–1.25 ± 0.53 mg/kg, respectively) can be ascribed to plant protection products or vine nutrition process. Values of the Cu and Zn concentration are higher than those reported in Brazil (79.87 ± 0.05 μg/100g grape berries - Cabernet Sauvignon and 31.56 ± 0.04 μg/100g grape berries - Merlot for Cu; 42.47 ± 0.17 μg/100g and 52.24 ± 0.74μg/100g for Zn) [33].

Although Pb occurs naturally in all plants, it has not been shown to play any essential role in their metabolism and its concentration at the level of 2–6 μg/g should be sufficient [25]. Pb has recently received much attention as a major metallic pollutant of the environment and as an element toxic to plants. Feteasca alba variety cultivated in Baia Sprie showed the highest Pb content (8.91 ± 1.50 mg/kg), other varieties from Baia Sprie and Baia Mare areas having similar Pb concentration (between 4.60 ± 0.64 and 6.80 ± 2.47 mg/kg). The varieties grown in Simleul Silvaniei recorded significantly lower concentration of Pb in grapes (Table 2). Overall, concentrations of Cd and Ni were detected in similar ranges in all three areas, though values tended to be higher in Baia Sprie and Simleul Silvaniei regions compared to Baia Mare. Cd is considered a non-essential element for metabolic processes; it is effectively absorbed by root and leaf systems and is also accumulated in soil organisms. There are evidences that an appreciable fraction of Cd is taken up passively by roots, but Cd is also absorbed metabolically [25]. There is no evidence of an essential role of Ni in plant metabolism, although several investigators suggested that Ni might be essential for plants. The essentiality of Ni for some biosynthesis of a number of bacteria has been proven. Also, its role in the nodulation of legumes and effects on the nitrification and mineralization of some OM was described [25]. Concentration of Co was higher in varieties from Baia Sprie area (3.26 ± 0.69–4.60 ± 2.54 mg/kg) as compared to Baia Mare (0.91 ± 0.06–1.26 ± 0.65 mg/kg) and Simleul Silvaniei (1.44 ± 0.29–1.93 ± 0.09 mg/kg). Co is cofactor of biosynthetic enzymatic activities essential for *Rhizobium*. Its content in plants is highly controlled by both soil factors and the ability of plants to absorb this metal [34]. In higher plants, absorption of Co by roots involves active transport [25]. Varieties from Baia Sprie had the highest As concentrations (0.83 ± 0.17–0.90 ± 0.07 mg/kg), followed by varieties from Baia Mare (0.29 ± 0.22–0.60 ± 0.30 mg/kg). No significant differences in Cr and Hg content were observed in all grape samples. The biochemistry of Hg is associated mainly with biological transformation of its compounds. However, it is not clear yet which processes are the most important in its cycling in the environment. In general, Hg content of plants is high when the Hg content of soils is also high, but this relation does not always hold. The results obtained are much higher than those reported in other studies [21,23,25,31].

#### *2.3. Metal Concentration in Must and Wine*

#### 2.3.1. Metal Concentration in must

Concentrations of Cu and Zn in must samples from Baia Sprie and Baia Mare areas exceeded the maximum permissible limit (M.P.L.) (10 mg/L), indicating a serious Cu and Zn pollution of the corresponding areas. Concentrations in the varieties cultivated in Simleul Silvaniei were below this threshold (Table 3). In grapevine nutrition, small quantities of Zn are taken from the soil (Zn is a trace mineral), so it is naturally present in must and wine. During alcoholic fermentation, part of the Zn precipitates due to the reducing environment and is accumulated in yeast.



*Molecules* **2020**, *25*, 750



0.1215 μg/L. LOQ for Hg: 0.1379 μg/L.

Concentrations of Pb in must were significantly higher in Baia Sprie area than in the other two areas. All varieties cultivated in Baia Sprie and grapes of Feteasca regala from Baia Mare slightly exceeded the M.P.L. (0.5 mg/L). Concentrations are higher than those reported for Brazilian grapes juice (0.07 ± 0.00 μg/100 mL grape juice - Cabernet Sauvignon; 0.11 ± 0.00 μg/100 mL grape juice - Merlot) [33], but lower for grapes juice originated from polluted and nonpolluted regions from Serbia (1.81 ± 0.15 mg/kg) [35]. Grapevine can accumulate small amounts of Pb (27–125 mg/kg), with an average of 58.2 mg/kg in grapes [36].

The highest concentrations of Cd in must were recorded in varieties cultivated in Baia Sprie, significantly higher than in Baia Mare. Grapes samples from Simleul Silvaniei had Cd concentrations below the limit of detection. Cd is a natural component of must as it originates from the grapes. During fermentation, up to 90% of Cd accumulates in yeast, thus wine contains 0001–0002 mg/L [36]. Interestingly, must of Feteasca alba variety cultivated in Baia Mare had remarkably higher concentration of Ni compared to other varieties or the same variety from other areas (Table 3). Our values are higher than those obtained for Brazilian grapes (0.40 ± 0.01 μg/100 mL grapes juice - Cabernet Sauvignon; 0.69 ± 0.00 μg/100 g mL grapes juice - Merlot) and lower than concentrations reported for grape berries juice from Serbia (2.16 ± 0.78 mg/kg and 1.77 ± 0.14 mg/kg, respectively) [35]. The level of Co, in must, is under the detection limit in all analyzed samples. As is usually present in must as a consequence of herbicides and insecticides used for grape production, processing factors, and must storage conditions [37]. Feteasca regala and Italian Riesling varieties from Baia Mare and Baia Sprie had significantly higher concentration of As in must samples (48.30 ± 1.27 μg/L (Feteasca regala); 46.35 ± 2.60 μg/L (Italian Riesling) from Baia Mare and 50.34 ± 2.75 μg/L(Feteasca regala); 49.87 ± 2.36 μg/L (Italian Riesling) from Baia Sprie) than Feteasca alba variety from the same areas (33.06 ± 1.58 μg/L and 35.68 ± 3.29 μg/L, respectively). Concentrations of As are below the M.P.L. in all tested must samples. Highest concentrations of Cr were recorded in must samples from Simleul Silvaniei for all three varieties, while Hg was detected in comparable amounts.

#### 2.3.2. Metal Concentration in Wine

Concentrations of Cu and Zn exceeded the M.P.L. under applicable law (1 mg/L for Cu and 5 mg/L for Zn) for varieties cultivated in Baia Sprie and Baia Mare and were below the M.P.L. in varieties from Simleul Silvaniei (Table 3). These concentrations are higher than those obtained in wine samples from different wine-producing areas of Romania: 403.92 μg/L (Cu) and 1183.32 μg/L (Zn) in Cabernet Sauvignon from Muntenia [38]; 886.31 μg/L (Cu) and 524.65 μg/L (Zn) from Muntenia, 289.52 μg/L (Cu) and 488.20 μg/L (Zn) from Dobrogea, and 642.60 μg/L (Cu) and 426.40 μg/L (Zn) from Moldova [26]. M.P.L. for Pb concentration in wine (0.15 mg/L) was exceeded in varieties from Baia Sprie and Baia Mare, the highest value being detected in Feteasca regala variety (0.38 ± 0.16 mg/L and 0.27 ± 0.02 mg/L, respectively). In other wine-producing regions, concentration of Pb was reported at 27.36 μg/L (Feteasca Neagra, Dealu-Mare) [39], 44.68 μg/L (Muntenia), 31.93 μg/L (Dobrogea), and 49.59 μg/L (Moldova) [26]. Concentrations of Cd in wine samples from Baia Sprie and Baia Mare were recorded within 0.02–0.06 mg/L, slightly above the M.P.L (0.01 mg/L); no statistically significant differences were observed between these values. In Simleul Silvaniei area, Cd concentrations were below the detection limit. Compared to our values, much lower Cd concentrations were reported for several red wine samples from Banat, Muntenia, Oltenia, and Dobrogea regions [39]. Concentration of Ni was statistically comparable in all three areas, values varying slightly between 0.02 mg/L (Simleul Silvaniei) and 0.08 mg/L (Baia Sprie). In comparison, in other Romanian wine-producing regions, similar values were reported in white wine samples but higher concentrations for red wines [26,39]. Co levels in wine samples were below the detection limit of the analytical method. Concentrations of As varied significantly amongst areas, for all three varieties, following the trend Baia Mare >Baia Sprie >Simleul Silvaniei, however, all values were below the M.P.L. imposed by law. In case of Cr, the trend was as follows: Baia Mare >Simleul Silvaniei >Baia Sprie. Concentrations of Hg were below the detection limit, except for Feteasca alba from Baia Mare (0.11 ± 0.02 μg/L).

#### *2.4. Pearson's Correlations Between the Content of the Investigated Elements From Soil, Plant Material, Must, and Wine*

The results of Pearson's correlation analysis revealed that there is a good negative correlation between metals contents in all plant parts and the distance from the "Romplumb" and "Cuprom" smelters, except for Cr and Ni in cane, Cr, Pb, and Ni in leave, Pb, As, Ni, and Co in grape, and Cd in must and wine (Table 4). Ni content in soil correlates positively with the distance. These results demonstrate that pollution resulted from metallurgical activities affect the heavy metal content of plant parts. Content of Cu, Zn, Cd, As, Pb and Hg in all plant parts decreased as the distance from the main pollution source increased, except for Ni content. Apparently, the Co smelter is not necessarily a dominant source of pollution for Pb, Co, Cr and As. These elements can be easily assimilated from soil naturally enriched with heavy metals and could come from combustion of fossil fuels in residential areas, heavy traffic, or some agricultural practices in rural zone [21,25].

**Table 4.** Pearson's correlation between the contents of the investigated element in plants parts and distance, between the contents of elemental in plants parts and related contents in soil, and between content in individual organs.


\*Correlation is significant at the 0.05 level (two-tailed); \*\*Correlation is significant at the 0.01 level (two-tailed).


**Table 4.** *Cont*.

\*Correlation is significant at the 0.05 level (two-tailed); \*\*Correlation is significant at the 0.01 level (two-tailed).

Significant positive correlation between metal level in plant and soil was detected in nearly all cases, while Co in grape, Pb in must, and Cr in grape and must showed significant negative correlations (Table 4). Although all elements in all samples, except for Cd, Ni, Co in grape and Cr in must, correlated positively with the metal content in roots, only the correlation of grape and root can be of interest as these organs reflect a real bioaccumulation [21].

Overall, the Pearson's correlation matrix for individual elements in soil, plant material, must, and wine showed a good positive correlation between contents of individual elements (Supplementary Table S2). Ni content in soil and Cd content in grape had negative correlation with other elements. Similar results regarding Ni behavior have been reported from Serbia [21]. The low correlation coefficients observed for Ni in soil and plant parts (except leaves) might indicate that this element

comes from different sources: Ni concentration in soil is impacted predominantly by geology, and the soil is mainly the source of Ni in plants parts. Leaves of grapevine from Baia Mare and Baia Sprie have captured Ni from atmospheres as well, originating from metallurgical activities. It is a known fact that above-ground plant parts assimilate elements from both soil and atmosphere, however, leaves are likely to be the most sensitive to air pollution.

#### *2.5. Translocation Factor (TF) and Mobility of the Element Content in the Soil-Grapevine-Wine System*

TF of the metals from the soil to the aerial parts of the plant represent an essential indicator of heavy metal mobility and translocation to the edible parts of the plant. Mobility ratio (MR) in *Vitis vinifera* L. was used to determine the ratio between the metal concentration in plant parts (canes, leaves and grapes) and the concentration levels of the acid-soluble metal faction in top soil. MR >1 indicates that the plants enrich these elements (accumulator), a ratio at around 1 indicates a rather indifferent behavior of the plant towards these elements (indicator) and a ratio clearly < 1 shows that the plant exclude these elements from uptake (excluder) [40].

Mean values of TF and MR indicated effective translocation of most elements in *Vitis vinifera* L. at all three sampling sites (Tables 5 and 6). Effective translocation of Ni (Feteasca alba), Co (Feteasca alba, Feteasca regala and Italian Riesling), As (Feteasca alba, Feteasca regala and Italian Riesling), Cr (Feteasca alba, Feteasca regala and Italian Riesling) occurs from soil to grapevine roots. From roots to canes, effective translocation was recorded for Pb (Feteasca alba, Feteasca regala and Italian Riesling), Cd (Feteasca alba, Feteasca regala and Italian Riesling), Ni (Italian Riesling), Co (Feteasca regala and Italian Riesling). From canes to leaves, translocation was recorded to Cu (Feteasca alba), Pb (Feteasca alba, Feteasca regala and Italian Riesling), Ni (Feteasca alba and Feteasca regala), Co (Feteasca alba), As (Feteasca alba, Feteasca regala and Italian Riesling) and Hg (Feteasca alba, Feteasca regala and Italian Riesling), while from grapes to must, effective translocation of Cu (Feteasca alba, Feteasca regala and Italian Riesling), Zn (Feteasca alba, Feteasca regala and Italian Riesling) and Cr (Feteasca alba, Feteasca regala and Italian Riesling) was detected. For most elements, translocation coefficient between grapes-cane, must-grapes, and wine-must had values lower than 1, indicating grapevine's specific mechanisms to block the accumulation of toxic metals in grapes [41–43]. The physico-chemical and biological processes that occur in the process of transformation the must into wine generates the reducing of the heavy metals concentrations, and this is demonstrated with the lower values of the analyzed metals in wine and in must as well from the values lower than 1 of the TFs [23] based on MR values, absorption of Cu, Zn, Pb from soil to roots, canes, leaves, grapes, must, and wine of all varieties of *Vitis vinifera* L. was not considerable (MR<1). In case of Cd (canes/soils), As (roots/soil and leaves/soil), Hg (canes/soil), MR value around 1 indicates that plants had an indifferent behavior against these elements. According to literature data, *Vitis vinifera* L. can be considerate as a bioaccumulator of Pb, Cu, and Zn [14,21]. Our results also demonstrated that *Vitis vinifera* L. is not a hyperaccumulator of Cu, Zn, Pb, Cd, Ni, Co, As, Cr and Hg (absorb metals above established background concentration).









#### *2.6. Combining Multielement Analysis of Must and Wine for Geographical Discrimination*

Elements like Mn, Cd, Li, Ba, Ca, Bi, Rb, Mg, Ag, Ni, Cr, Sr, Zn, Rb and Fe showed a high discriminatory power for geographic origin of Romanian wine, but additional new elements (Hg, Ag, As, Al, Tl, U), metal ratios (Ca/Sr and K/Rb) and 207Pb/ 206Pb, 208Pb/ 206Pb, 204Pb/ 206Pb, 87Sr/ 86Sr isotope ratios have been investigated in order to identify new tracers for geographical traceability of Romanian wines [24,26,44].

This is the first study to assess the geographic fingerprinting of wine and must samples from a polluted area (Baia Mare and Baia Sprie). The analyzed wine samples showed high concentration of elements, but not exceeding the maximum levels recommended by International Organisation of Vine and Wine (OIV 2016), except for Cu, Zn, Pb and Cd in Baia Mare and Baia Sprie. In Simleul Silvaniei, the high concentration of some elements is mostly derived from agricultural practices, fertilizers, and technological winemaking processes. Multivariate chemometric method was applied for the differentiation of must and wine intro groups based on their geographic origin. Linear discriminant analysis (LDA) was used to identify significant tracers for classification to the geographical discrimination of the wine samples.

Based on the elemental contents, cross-validation technique provided an 88.09% and 84.87% percentage of predicted membership according to the must and wine geographic origin, respectively (Supplementary Figures S1 and S3). The linear correction revealed acceptable scores for the two defined discriminant factors (F1 = 73.09%, F2 = 15.01% for must and F1 = 62.36%, F2 = 22.50% for wine). F1 mainly separates Baia Mare and Baia Sprie areas from Simleul Silvaniei and F2 separates Simleul Silvaniei from Baia Mare and Baia Sprie (Supplementary Figure S2). Among the investigated parameters, Cr, Hg, As, Cu, Zn, Pb, Ni and Cd was identified as the most significant for geographic differentiation of the must and wine from Baia Mare, Baia Sprie, and Simleul Silvaniei areas. The technique of cross-validation was applied during the set validation and the proposed model appears to be a promising chemometric approach for precise classification of wines according to their geographical origin. Thus, in both cases, the geographical regions were correctly classified with percentage between 52% and 71%.

#### *2.7. Cluster Analysis*

The hierarchical dendrogram for polluted sites based on elements content in sol material (Supplementary Figure S5) showed two primary clusters of the contaminated locations. The first cluster is formed of sites located in Simleul Silvaniei area, while the second one is formed of sites from Baia Mare and Baia Sprie. In terms of measure interval, the difference between the two primary clusters was significant, which suggests higher soil pollution in Baia Mare and Baia Sprie compared to Simleul Silvaniei. Both primary clusters were further divided into several new subclusters. However, the differentiation between the areas from Baia Mare and Baia Sprie was more significant than Simleul Silvaniei area. The position of an isolated subcluster which belongs to the Baia Mare area suggested that this area is the most polluted one. The dendrogram of elements in vineyard soil (Supplementary Figure S6) showed two main clusters (one isolated for As and other for the rest of elements) and numerous subclusters. The difference between primary clusters was significant, which confirmed the previous conclusion that the source of As content in soils is of geological origin, whereas the concentrations of other metals in soil are also influenced by atmospheric pollution. This was particularly obvious in the case of Cu, Pb, Zn and Cd. Similar conclusions can be formulated from analysis of the dendrogram based on element contents in grapevine roots (Supplementary Figure S7), that indicated one cluster for Ni and another cluster for the rest of elements, as well as numerous different subclusters. The dendrogram of elements in grapevine canes, leaves and grapes (Supplementary Figures S8–S10) showed two main cluster: one isolated for Hg (canes dendrogram), As (leaves dendrogram), and Cd (grapes dendrogram) and another for the rest of elements; and several different subclusters. These results also demonstrated the two possible sources of the investigated elements in these organs: soil or atmosphere. The hierarchical dendrogram for must and wine based on elements content

(Supplementary Figures S11 and S12) showed two primary clusters. For must, first cluster is formed by Zn, Cu, Hg, Cd, Pb, Co and As and the second cluster is formed by Cr and Ni. For wine, first cluster is formed by Pb, Zn, Cd, Cu, Ni and the second cluster is formed by Hg, Co, Cr, As. The hierarchical dendrogram for the elements in the upper organs of grapevine (Supplementary Figure S13) also showed two main clusters: one cluster formed by Co, Ni, Hg, Cd, (grapes), Hg (canes), As (leave) and other for the rest of elements, in canes, leaves and grapes, as well as numerous different subclusters which demonstrated well a fine structure with two possible sources for the investigated elements: soil or atmosphere. The grouping of the elements confirmed that the Co, Ni, Hg, Cd, As concentrations of soil are the main source of Co, Ni, Hg, Cd, As content in the upper organs and the influence of atmospheric pollution is the highest for the group consisting of: Zn grape, Cr cane, Co leave, Cr, grape, Cr leave, As cane, that are placed furthest from the primary cluster. The combination of methods used in this study for data analysis, such as the calculation of TFs, MRs, Pearson's correlation study, and hierarchical cluster analysis, provided a very valuable information that made feasible a multi-aspect construction of the grapevine study and can be recommended for any similar investigation.

#### **3. Materials and Methods**

#### *3.1. Description of the Sampling Area*

The present study was conducted in Baia Mare and Baia Sprie area, one of the important mining districts in Romania. The main mining activities previously developed in the area considered of nonferrous sulfidic ore extraction and processing, aiming to obtain concentrated of Pb, Cu, Zn and precious metals. After 2006, the metallurgical industry from Baia Mare and Baia Sprie has considerably diminished its activity by closing or reducing its production capacity.

Baia Mare depression is a contact depression the interposes between the Somes, ana Plain and the Carpathian Mountains as a lower morphological unit, from the surrounding areas, presenting a waved surface, characterized by a convergent system of valleys and interfluves. It was formed due to the tertiary tectonic movement that took to the fragmentation and sinking of the crystalline in the Northwest part of Transylvania, as well as due to the volcanic chain of the Gutin-Oas, Mountains. The metropolitan area of Baia Mare is in the NW of Romania, in a hilly region, at an altitude of 220 m above sea level, covering an area of 1250 km<sup>2</sup> and having a population of more than 200.000 residents.

The Simleul Silvaniei vineyard is located in the northwest of Romania and is delimited by the Apuseni Carpathians on the south, the Somes, an Plateau on the east and the Somes, an Plain on the northwest, which is known geographically under the name of Silvaniei Hills. The altitude of this depression decreases from 500 m, in the accumulation area under the mountain, at 350–300 m, located in the wide part between the Măgura S, imleului and the Plopis, Mountains. Because of its position is among the northernmost vineyard in Romania. The climate of Baia Mare, Baia Sprie and S, imleul Silvaniei area falls in both moderate continental and the mountain climate categories [45].

#### *3.2. Description of the soil types*

According to the Romanian Soil Taxonomic Classification [46] in the investigated areas there were found: eutricambosol, typical luvosol, stagnic luvosol, gleyic luvosol, and aluviosols. Vegetation characteristic of eutricambosol soils was represented by forests partly replaced by pastures and meadows. Eutricambosols are moderate acidic with a slight difference on soil profile. Humus content is relatively high in the organic horizon (2.76–4.44%) [46]. Luvisols were represented by typical stagnic and gleyic luvosol types. They appear on a small area near metallurgical plant and are prevalent in the southern extension of the investigated areas. These soils are developed on the low plains and poorly drained terrains. Typical luvosol was present on large areas, being covered by orchards and grasslands. The Ao horizon has grey colour. The colors of the Bt horizon vary from red to brown. Soil profile was as follows: Ao-Bt-C. Stagnic and gleyic luvosol types were poor in nutrients and humus and had low natural fertility being covered by natural grasslands. Soil profile was as follows:

Ao-Ea(El)-E/B-Bt-C. The Ao horizon was 15 cm thick, the brown-grey color indicating a low content of humus. The structure was granular; the texture ranging from clay loamy to clay. Aluviosols were presented only in the western proximity of metallurgical plant and were consisted of an Ao horizon of 40 cm, which on top of C horizon of alluvial deposits [46,47].

#### *3.3. Sample Collection and Processing*

Soil, cane, and leave samples of grapevine were collected from Baia Mare, Baia Sprie and surrounding areas (Simleul Silvaniei) (Figure 1) during the vegetation period in May 2012. Soil samples were collected at the depth of 0–20, 20–40, 40–60 and 60–80 cm at the vineyard. Grapes of Feteasca alba, Feteasca regala, and Italian Riesling varieties were sampled one week before harvesting in August 2012. Roots (diameters <2.5 mm and >2.5 mm), canes (50–70 cane pieces of 25 cm), and leaves (50–70 fully-developed leaves from the middle part of the one-year old cane) were also collected. After removing damaged plant materials, samples were placed in sealed plastic bags and were immediately transported to the laboratory. Plant materials and soil samples were carefully processed to avoid chemical and physical interactions and analyzed by Inductively Coupled Plasma Mass Spectrometry (ICP-MS) Waltham, Massachusetts, SUA (see the Supplementary Materials).

**Figure 1.** Map of the Mining and Smelting Complex Baia Mare (Northwest Romania) with the sampling points.

#### *3.4. Soil Sample Preparation*

The soil samples (100 samples) were dried, homogenized and then passed through a 20-mesh sieve to obtain very fine particles. The method for microwave digestion using a Milestone START D Microwave Digestion System (Sorisole, Italy) was optimized in a previous work [22]: 0.25 g soil, 9 mL 65% HNO3, 3 mL concentrate HF and 2 mL concentrated HCl were placed in a clean Teflon digestion vessel. The vessel was closed tightly and placed in the microwave. The digestion was carried out with the program described in Supplementary Table S3.

#### *3.5. Plant Material Samples (Roots, Canes and Leaves) Preparation*

The plant material samples (75 samples of roots, 113 samples of canes and 140 samples of leaves) were thoroughly washed with tap water followed by ultra-pure water using Milli-Q Integral ultrapure water-Type 1 (Darmstadt, Germany), after washing was oven-dried at 80 ◦C to constant weight using a FD 53 Binder (Darmstadt, Germany). The dried samples were ground using a Retsch 110 automatic mill (Darmstadt, Germany), passed through a 2 mm sieve to obtain very fine particles. The method for

microwave digestion using a Milestone START D Microwave Digestion System (Sorisole, Italy) was optimized in a previous work [23]: 1 g sample of plant material, 7 mL 65% HNO3 and 2 mL H2O2 were placed in a clean Teflon digestion vessel. The vessel was closed tightly and placed in the microwave. The digestion was carried out with the program described in Supplementary Table S3.

#### *3.6. Grape Juice Sample Preparation*

Grape samples (100–110 kg/cultivar) were collected from each cultivar from 70 vines. The grapes placed in the top, middle and lower third of each vine and grapes were exposed to sun and shade [22]. In this way can achieve better homogenization of sample grapes. Feteasca regala (three samples), Feteasca alba (three samples), Italian Riesling (three samples) grape juices (must) were cold pressed manually. Before the analysis, each juice samples (50 mL) were diluted in different proportions using ultrapure water. All samples were taken in triplicates from the defined experimental plot of which had a size of 5 ha.

#### *3.7. Microvinification Process*

The samples of grapes were destemmed and crushed, then transferred to a microfermentor (50 L) cylindrical glass container, covered with aluminium foil to limit the effect of the light over the must) equipped with a fermentation airlock. Fermentation took place at 22–24 ◦C and humidity 55–60%. Afterwards wine was clarified by means of bentonite (40 g/L 1:10 dilution) and combined with SO2 up to 100 g/L. Then wines were allowed to cool for thirty days at −5 ◦C for cold stabilization [23]. Then wine samples were stored in glass bottles at 5–6 ◦C until the analyses. Average data from three vinifications per cultivar are reported [23].

#### *3.8. Wine Sample*

The wine samples were taken from freshly opened bottles and prepared by a specific organic matter digestion. 2.5 mL of wine were weighed inside Teflon digestion vessels and 2.5 mL concentrated HNO3 added. Teflon digestion vessels were previously cleaned in nitric solution to avoid cross-contamination. The vessels already capped were placed in a microwave oven followed by the application of the program described in Supplementary Table S3, optimized in a previous work [23]. After cooling to ambient temperature, the microwave oven was opened and the content was quantitatively transferred into a 50 mL volumetric flask and brought to the volume with ultra-pure water. All the elements were measured from these extraction solutions by ICP-MS (Waltham, Massachusetts, SUA).

#### *3.9. Inductively Coupled Plasma Mass Spectrometer (ICP-MS) Analysis*

Analytical measurements were performed using an inductively coupled plasma mass spectrometer (iCAP Q ICP-MS Thermo Fisher Scientific, Waltham, Massachusetts, SUA) equipped with an ASX-520 autosampler, a micro-concentric nebulizer, nickel cones and peristaltic sample delivery pump, running a quantitative analysis mode. Each sample was analyzed in duplicate and each analysis consisted of seven replicates. The gaseous argon and helium used to form the plasma in the ICP-MS was of purity 6.0 (Messer – Gases for Life, Austria). The heavy metals were measured by using a multi-element analysis after appropriate dilution using an external and standard calibration. The calibration was performed using XXICertiPUR multielement standard, and from individual standard solution of Hg. The working standards and the control samples were prepared daily from the intermediate standards that were prepared from the stock solution. The intermediate solutions stored in polyethylene bottles and glassware were cleaned by soaking in 10% *v*/*v* HNO3 for 24 h and rinsing at least ten rimes with ultrapure water (Milli-Q Integral ultrapure water-Type 1). The accuracy of the methods was evaluated by replicate analyses of fortified samples (10 μL–10 mL concentrations) and the obtained values ranged between 0.8–13.1%, depending on the element. The global recovery for each element was estimated and the obtained values were between 84.6–100.9%.

For quality control purpose, blanks and triplicates samples (*n* = 3) we analyzed during the procedure. The variation coefficient was under 5% and detection limits (ppb) were determined by the calibration curve method. Limit of detection (LoD) and Limit of quantification (LoQ) limits were calculated according to the next mathematical formulas: LoD = 3×SD/s and LoQ = 10×SD/s (SD = estimation of the standard deviation of the regression line; s = slope of the calibration curve) (Supplementary Table S4). The recovery assays for the must and wine sample of 5 μL concentration, for three replicates of this level of concentration (*n* = 3) gave the average recovery R % between 87.32% and 100.26%. The recovery for the soil and plant material samples of 5 μL concentration, for three replicates of this level of concentration (*n* = 3) gave the average recovery R % between 83.41% and 109.02%. Optimum instrumental conditions for ICP-MS measurement are summarized in Supplementary Table S3. The calibration standards were prepared from the multielement standard solution, ICP Multi Element Standard Solution XXI CertiPUR, in five concentration ranges 2.5, 5, 10, 25 and 50 μL.

#### *3.10. The Determination of pH, Electrical Conductivity (EC) and Organic Matter (OM)*

The pH and EC of soil samples (soil/distilled water = 1:2.5) were measured using pH meter Jenway, 3510, Keison (Chelmsford, UK) and an Electrical Conductivity (EC) meter Jenway, 3510, Keison (Chelmsford, UK), respectively. The organic matter (OM) was determined by loss-on-ignition method at 550 ◦C [21].

#### *3.11. Reagents and Solutions*

High purity ICP Multi-element Standard Solution XXI CertiPUR obtained from Merck (Darmstadt, Germany) was used for the calibration curve in the quantitative analysis. HNO3, concentrated HF and HCl (reagent grade from Merck, Darmstadt, Germany) and ultra-pure water (maximum resistivity of 18.2 M <sup>×</sup> cm-1, Milli-Q Integral ultrapure water-Type 1) were used for sample preparation.

#### *3.12. Statistical Analysis*

Average and standard deviation were calculated, and data were interpreted with the analysis of variance (ANOVA) and the average separation was performed with the Duncan test at *p* ≤ 0.005. Pearson's correlation coefficient was calculated using SPSS Version 24 (SPSS Inc., Chicago, IL, USA), Excel 2016 (Microsoft, New York, NY, USA) and Addinsoft version 15.5.03.3707 (Microsoft, New York, NY, USA. Value higher than 0.5 indicate a strong correlation between analyzed varieties, a positive correlation between two parameters shows that both parameters increased, and a negative correlation indicates that a parameter increased while the second one decreased and vice-versa. Linear discriminant analysis (LDA) was performed to separate the wines by region and to identify the markers with a significant discrimination value (variables with Wilk's lambda near zero, *p* values <0.005 and higher F coefficients), using Microsoft Excel 2016 and XLSTAT Addinsoft version 15.5.03.3707. By cross-validation, we established the optimal number of parameters required to obtain a robust model.

Trace metal TF in grapevine was determined by the equation (TFr-s = Croots/Csoils; TFc-r = Ccanes/Croots; TFl-c = Cleaves/Ccanes; TFm-c = Cmust/Ccanes; TFw-m = Cwine/Cmust as the ratio between roots-soil; canes-roots; leaves-canes; must-canes, and wine-must. TF > 1 indicates that grapevine translocates metals effectively from soil to plants parts [43]. The MR between the metal concentration in plant parts (Cplant, mg/kg) and concentration in the top-soil (Csoil–m, mg/kg) was determined according to the equation MR = Cplant/Csoil-m. MR > 1 indicates effective metal translocation from soil to plants parts.

#### **4. Conclusions**

All organs and products of *Vitis vinifera* L., except for grapes, must, and wine, provide numerous pieces of reliable information for efficient biomonitoring. Obtained data showed a very low environmental quality of the ecosystem in Baia Mare, Baia Sprie, and their surrounding areas. Furthermore, the content of most elements in plant parts is affected by airborne pollution which

comes from nearby metallurgical activities, i.e., from the Cu smelter, whereas geology contributes predominately to the Ni content. Also, these results suggest that the Cu smelter is not necessarily a dominant source of pollution by As and Hg.

The most abundant elements in all plants, soil samples, must, and wine from Baia Mare and Baia Sprie areas were Cu and Zn, except for grape samples. Apparently, the investigated grapevine cultivar poses some specific means for a strong protection of grapes from high concentrations of heavy metals, while tolerates considerable amounts of heavy metals (Cu, Zn, Hg, As) in other tissues, especially in root tissue. This means that the *Vitis vinifera* L. cultivated in Baia Mare and Baia Sprie areas may have developed a wide range of cellular mechanisms that are highly effective in heavy metal detoxification and tolerance to heavy-metal-induced stress, including different tactics of restriction of metal uptake from soil as well as the retention of assimilated metals in the root tissue. Except of sporadic incidences, there were no visible symptoms of phytotoxic effects of metals, even though many of the grapevines were growing in highly polluted soils. Planting of the *Vitis vinifera* L. can be recommended in all kinds of soils that are severely polluted with heavy metals because it is a suitable candidate for phytostabilization. The plants of this climber species may also be useful as a vegetation protection barrier from considerable atmospheric pollution. At the same time, berries are safe for consumption to a large degree, which is a great advantage of this species.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1420-3049/25/3/750/s1, The physical properties of the soils samples; Table S1. The physical properties of the soils samples (Mean ± standard deviation) (*n* = 3); Table S2. Pearson's correlation matrix for investigated elemental in sol, plant material, must and wine; Figure S1. Correlation between analyzed parameters and the factors in discriminant analysis of must geographic origin; Figure S2. Differentiation of must according to geographic origin based on elements content; Figure S3. Correlation between analyzed parameters and the factors in discriminant analysis of wine geographic origin; Figure S4. Differentiation of wine according to geographic origin based on elements content; Figure S5. Hierarchical dendrogram for polluted sites based on element contents in soils; Figure S6. Hierarchical dendrogram for elements in vineyard soil; Figure S7. Hierarchical dendrogram for elements in grapevine roots; Figure S8. Hierarchical dendrogram for elements in grapevine canes; Figure S9. Hierarchical dendrogram for elements in grapevine leaves; Figure S10. Hierarchical dendrogram for elements in grapevine grapes; Figure S11. Hierarchical dendrogram for elements in grapevine must; Figure S12. Hierarchical dendrogram for elements in grapevine wine; Figure S13. Hierarchical dendrogram for elements in grapevine upper organs; Table S3. The program of the microwave oven Milestone START D Microwave Digestion System; Table S4. LoD, LoQ, BEC and r2 of the calibration for each element; Table S5. Instrumental (a) and data acquisition (b) parameters of ICP-MS.

**Author Contributions:** F.D.B. and C.I.B. conceived and designed the experiments; F.D.B. performed the sample collection and processing, the determination of pH, electrical conductivity and organic matter and wrote the first draft of the manuscript. C.I.B. and R.C. contributed to statistical analysis and manuscript revision. A.B. contributes to data analysis and revised the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** The publication was supported by funds from the Ministry of Research and Innovation through Program 1 - Development of the National Research and Development System, Subprogram 1.2 - Institutional Performance - Projects for Financing the Excellence in CDI, Contract no. 37PFE/06.11.2018.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Analysis of Pollution in High Voltage Insulators via Laser-Induced Breakdown Spectroscopy**

#### **Xinwei Wang 1, Shan Lu 1, Tianzheng Wang 1, Xinran Qin 2, Xilin Wang 2,\* and Zhidong Jia <sup>2</sup>**


Academic Editors: Clinio Locatelli, Marcello Locatelli and Dora Melucci Received: 13 December 2019; Accepted: 10 February 2020; Published: 13 February 2020

**Abstract:** Surface pollution deposition in a high voltage surface can reduce the surface flashover voltage, which is considered to be a serious accident in the transmission of electric power for the high conductivity of pollution in wet weather, such as rain or fog. Accordingly, a rapid and accurate online pollution detection method is of great importance for monitoring the safe status of transmission lines. Usually, to detect the equivalent salt deposit density (ESDD) and non-soluble deposit density (NSDD), the pollution should be collected when power cut off and bring back to lab, time-consuming, low accuracy and unable to meet the online detection. Laser-induced breakdown spectroscopy (LIBS) shows the highest potential for achieving online pollution detection, but its application in high voltage electrical engineering has only just begun to be examined. In this study, a LIBS method for quantitatively detecting the compositions of pollutions on the insulators was investigated, and the spectral characteristics of a natural pollution sample were examined. The energy spectra and LIBS analysis results were compared. LIBS was shown to detect pollution elements that were not detected by conventional energy spectroscopy and had an improved capacity to determine pollution composition. Furthermore, the effects of parameters, such as laser energy intensity and delay time, were investigated for artificial pollutions. Increasing the laser energy intensity and selecting a suitable delay time could enhance the precision and relative spectral intensities of the elements. Additionally, reducing the particle size and increasing the density achieved the same results.

**Keywords:** laser-induced breakdown spectroscopy; surface pollution; high voltage insulators; quantitatively analysis

#### **1. Introduction**

The insulators were key equipment in transmission lines, in order to mechanically support conductor and give enough insulation space between conductor and tower. After being in operation for in a transmission line, an insulator (ceramic, glass or composite insulator) can accumulate a thick layer of pollutants on its surface due to different environmental factors. Under dry conditions, pollution was not harmful and had little effect on the safe service. However, soluble pollutants can be dissolved in water, forming a conductive water film on the surface of an insulator; this process results in the formation of conductive channels on the surface of the insulator, and in turn, reduces the pollution flashover voltage (PFV), thereby causing partial discharge, arc and even flash-over incidents [1]. Methods for detecting the pollution characteristics and pollution level of insulators have been studied for a long time. The Working Group 04 of Study Committee 33 (Over-voltage and Insulation Coordination) of the International Council on Large Electric Systems has recommended

five methods for quantitatively characterizing pollution levels, including the equivalent salt deposit density (ESDD), surface conductivity, leakage current, PFV and pollution flashover gradient.

Pollution composition is complex and differs between environments. In nature, soluble pollutions are primarily conductive electrolytes, such as NaCl, KCl, CaSO4, CaCl2, Na2SO4, NaNO3 and KNO3; the main insoluble pollutions include SiO2, C, Al2O3, MgSO4, Fe2O3 and CaO [2,3]. Researchers have found that the pollution levels measured by ESDD differ from the actual values to a certain extent. As a result, the PFVs of artificial pollutions are lower than those of natural pollutions with the same ESDD. The PFV of the artificial pollution CaSO4 is higher than that of the artificial pollution NaCl for the same ESDD. Additionally, for an artificial pollution mixture of NaCl and CaSO4, the higher the CaSO4 content, the higher the PFV is [4–7].

Currently, researchers also employ other indirect methods (e.g., light, sound and electricity) to determine the pollution levels. Hyperspectral imaging, microwave radiation theory, infrared and visible light information fusion, ultraviolet sensors, light detection sensors and acoustic emission technology have been employed to establish insulator pollution level prediction models [8–15]. With respect to the direct detection of pollution composition, aside from commonly used material composition analysis methods (e.g., ion emission spectroscopy techniques, including ion chromatography, X-ray diffraction and inductive coupling), very few researchers have examined online detection methods for insulator pollution composition. However, the compositional distribution of pollutions on surface of an insulator is often complex and heterogeneous. These factors present difficulties for evaluating pollution levels by indirect methods. Additionally, research results have demonstrated that pollution composition and material characteristics can affect the pollution flashover process, and may cause excessive or deficient insulation in insulation design [16–18]. The PFV is not only related to soluble salt composition, but also affected by insoluble substances in different mixtures [19].

To improve the accuracy and application of LIBS. The researchers studied various sample preparation techniques, such as dilution and using binding material, etc. By milling [20] and grinding, the particle size is reduced and the surface area is increased to make the sample more uniform. The smaller the particle size, the easier it is to evaporate and atomize in the plasma [21].

Laser-induced breakdown spectroscopy (LIBS) is a qualitative and quantitative analytical method based on pulse laser technology that examines the plasma atomic emission spectrum after exciting the sample [22], and it had higher sensitivity for light elements detection (H,Li,C,Si etc.), compared to EDS (or EDX) technique [23,24]. Currently, owing to the rapid development of this technique, the use of LIBS is widespread in the theoretical and experimental research of many fields, such as those of mineral products, archaeology, biomedicine and aerospace exploration [25–29]. In particular, LIBS is currently the only feasible technique in fields that require remote elemental analysis [26]. We have [30,31] evaluated the feasibility of using LIBS to achieve rapid, accurate, online monitoring of the ageing performance of silicone rubber and to determine the components (C, O, Fe and Si) that are closely related to the ageing state of silicone rubber. Combined with XPS technology, the linear calibration curves of these components were established. Based on the variation trend of element spectral intensity with depth, the depth of aging layer was obtained. However, compared with silicone rubber, pollution composition is more complex and varied. Therefore, when studying pollutions using LIBS, it is necessary to consider the effects of the properties of the pollutions and optimize the system parameters. In this study, the effects of various factors on the LIBS spectra of natural and artificial pollutions are examined and optimized system parameters are proposed.

#### **2. Results and Discussion**

#### *2.1. Microanalysis and LIBS Testing of Pollutions Sampled*

Figure 1 shows SEM images of the natural pollutions on the surface of the insulator at two randomly selected analytical points. As demonstrated in Figure 1, the pollutions at sampling point 1 were densely distributed and exhibited layer-by-layer stacking. In contrast, the pollutions at sampling

point 2 were loosely arranged, and there were relatively large spaces between the pollutions. The process by which natural pollutions were adhered to the surface of an insulator is affected by the air flow in the environment. The uneven adhesive forces between particles and the surface of an insulator, due to the ageing of the insulator and the random interactions between particles, can result in uneven adherence of pollutions on the surfaces of adjacent insulators.

**Figure 1.** Scanning electron microscopy (SEM) images (5000×) of the natural pollutions on the surface of the insulator (**a**) Sampling point 1, (**b**) sampling point 2.

The EDS detector of the SEM was used to analyze the elements on the surface of the insulator. Table 1 summarizes the results. Very few elements were detected by EDS, and minimal Cl was detected. Natural pollutions often contain NaCl and KCl, which significantly affect pollution flashovers. The NaCl and KCl on the surface of the insulator may have been eliminated by dissolution and scouring as a result of dampening and rainfall. The EDS detector only analyzed the surface composition of the sample and consequently failed to detect the distributions of other common elements. Titanium ore is in the area of insulator operation, so there is high concentration of Ti in the pollution. Therefore, other methods were needed to determine the composition of pollutions on the surface of the insulator.

**Table 1.** Energy dispersive X-ray spectroscopy (EDS) analysis results for the composition of natural pollutions on the surface of the insulator.


Figure 2 showed the LIBS spectrum of the natural pollutions. Table 2 showed the wavelength of typical spectral lines in Figure 2. In Figure 2, the wavelengths of the abscissa correspond to the emission intensities of various elements. Each element has multiple emission lines. In testing, a characteristic wavelength should be selected, and the type of element and relative spectral intensity, corresponding to the characteristic wavelength, should be determined [32]. Spectral intensities reflect the composition of the sample tested. As shown in Figure 2, Si, Ca, Al, C and Na had relatively high intensity, and this indicates that, agreeing well with the EDS area scan results, the natural pollutants had relatively high contents of these elements.

**Figure 2.** LIBS spectrum of the natural pollutions.



Trace amounts of Na and Mg were detected in the samples tested by LIBS, while the EDS area scans of the sample, Na and Mg were not detected in the ablation pits of the silicone rubber. Therefore, LIBS can not only achieve rapid, online detection of elements, but also help further reduce the detection limit of current composition testing and improve the accuracy of quantitative/qualitative compositional analysis.

#### *2.2. E*ff*ects of Single-Pulse Laser Energy on the LIBS Signal*

In LIBS, the depth of ablation craters depends on many factors, such as laser energy, ablation duration and material characteristics. The single-pulse laser energy has an impact on the ablation of pollutions on the surfaces of insulators. Ideally, a single-pulse laser beam only ablates the pollutions on the surface of an insulator but not the surface of the insulator itself. Figure 3 shows SEM images (200×) of the laser-ablated samples (labelled top and bottom). As demonstrated in Figure 3a,b, a focused laser beam produced an ellipsoidal ablation pit on the surface of each sample, which was related to the morphology of the focused laser beam. The sample bottom was taken from the surface of a silicone rubber insulator that had aged as a result of being in service for an extended period of time, and cracks differing in size were distributed on its surface. To further analyze the ablation effects of a single-pulse laser on the natural pollutions on the surface of the insulator, EDS area scans were performed on the natural pollutions and the ablation pits on the surfaces of the samples top and bottom to analyze the elemental compositions. The results showed that the typical characteristic elements (e.g., Na, Mg and Ti) were not detected in these samples by EDS. This observation suggests that the LIBS testing, with a single- pulse laser beam with an output energy of 110 mJ, was able to penetrate the relatively thin (micro-sized) pollution layers and ablated the pollutions at the point of action into laser plasma, thereby, exposing the substrate of the insulator.

**Figure 3.** Distribution of the natural pollutions on the surface of the insulator (**a**) SEM image of the entire ablated sample top, (**b**) SEM image of the entire ablated sample bottom.

A laser energy increases within a certain range, the energy absorbed per unit target surface area increases, resulting in an increase in the spectral intensity of the sample. Once the increase in laser energy outside this range may result in self-absorption of or matrix effects on elements, which in turn, results in a decrease in intensity. In the experiment, artificial pollutions were prepared to determine the spectral intensities under various laser energies within a reasonable range. Figure 4. shows partial LIB spectra, obtained under various strengths of laser energies. As demonstrated in Figure 4, as the laser energy increased, the spectral intensities corresponding to different wavelengths increased by varying degrees. The spectral intensity of Al corresponding to a wavelength of 396.592 nm saturated prematurely. Therefore, while higher laser energy may improve the spectral intensity, extremely high laser energy outranging a certain range may interfere with the experiment.

**Figure 4.** LIBS spectra within a certain band range under various laser energies.

This work done with the insulator that has been exposed to the elements. The LIBS method was described as follows: Five points on the surface of each sample were randomly selected. Each point was subjected to five continuous laser treatments. Figures 5 and 6 show the effects of laser energy density on the spectral intensity, and relative standard deviation (RSD) of various elements tested, respectively. The laser energy intensity was obtained by dividing the laser energy by the spot area. The diameter of laser focusing on sample surface was 0.8 mm. The spectral line intensity increased as the pulsed laser energy intensity increased. As demonstrated in Figure 6, as the spectral intensity increased, the RSDs of almost all the elements gradually decreased, suggesting that increasing laser energy intensity could effectively improve the repeatability of results. The RSD is related to the concentration and spectral line intensities of the sample and is affected by the spectral analysis conditions and instrument performance.

**Figure 5.** Effects of laser energy on the spectral intensities of the elements tested

**Figure 6.** Relationship between the relative spectral intensity and measurement repeatability of the elements tested.

Additionally, an increase in laser energy intensity provided sufficient excitation energy for certain elements, causing intensity saturation or self-absorption effects and consequently decreasing the peak values. Meanwhile, owing to matrix effects, the increase in laser energy significantly interfered with the spectral information of other elements, leading to negative effects. Based on the SEM results for the pollutions subjected to LIBS testing, the laser energy was adjusted to approximately 80 mJ, corresponding to a laser ablation density of 3.814 <sup>×</sup> <sup>10</sup><sup>10</sup> Watts/cm2. The ablation effects of the adjusted laser energy on the surface of the composite insulator were comparatively analyzed.

#### *2.3. Selection of Delay Time*

Figure 7 shows the trends of the spectral intensities within the same band range with the delay time. As demonstrated in Figure 7, as the delay time increased, the normalized spectral intensity corresponding to each wavelength significantly decreased. Using the average relative spectral intensity at a delay time of 0.5 μs as the baseline, the normalized relative spectral intensities were calculated by dividing the average relative spectral intensities at other delay times by the baseline. Figure 8 shows the results, the experimental results show that the trends of the spectral intensities of each element corresponding to various wavelengths were similar. Hence, only the spectral intensity of one element, corresponding to one wavelength, was selected for analysis.

**Figure 7.** Changes in LIBS spectra within a certain band range with delay time.

**Figure 8.** Effects of delay time on the relative spectral intensity of each element.

As demonstrated in Figure 8, the continuous background spectral process was not complete at a delay time of 0.5 μs. As the delay time increased, the spectral intensity of each element considerably decreased. Additionally, as the delay time increased, the RSD for Ca first slowly increased, and then gradually stabilized as shown in Figure 9. During the plasma cooling process, the collisions between ions and electrons continuously weakened, and consequently, the luminous intensities of energy released from the collisions and received by the spectrometer continuously decreased. In particular, as the delay time increased from 1 to 9 μs, the normalization ratio for Na fluctuated in the range of 0.3–0.45 because Na, being an alkali metal element prone to ionization, was completely ionized within 1 μs. As a result, as the measurement delay time increased, the number of Na ions received by the system decreased, resulting in a decrease in the measurement accuracy.

**Figure 9.** Effects of delay time on the repeatability of measurements on the elements tested.

Considering the relationships among the spectral intensity, RSD and delay time, a delay time ranging from 2 to 4 μs was selected as the optimum delay time range that led to a normalization ratio greater than 0.4 and an RSD less than 20%. A gate-width delay time of 3 μs was used in the subsequent experiment. When analyzing a particular element, a delay time range that leads to a normalization ratio greater than 0.5 and a minimum RSD should be selected.

#### *2.4. E*ff*ects of Pollution Particle size and Density on LIBS Signal*

Pollutions on the surface of an insulator in operation have complex and varied compositions (as shown in Figure 1). Pollution particles vary in size between different locations, and the gap density varies between pollutions. Inconsistent particle sizes and densities can both affect LIBS spectra.

First, the effects of pollution particle size on the LIBS spectral signal were studied. A Malvern Mastersizer 2000 laser particle-size analyzer was used to measure the particle size of the NaCl samples and kaolin clay [33]. A wet method was employed, and ethanol was used to dissolve the samples. Table 3 summarized the particle-size test results.


Distribution/50% 342.3 240.763 140.694 60.914 5.346


Based on the total area of XP-70, the mass of NaCl in each sample was determined. Each sample was compressed using a compression machine and subsequently subjected to LIBS testing. Figure 11 shows the changes in the relative spectral intensities of the Na in NaCl samples differing in particle size with NaCl concentration. When the NaCl particle size remained constant, the average relative spectral intensity of the two spectral lines of Na first increased and then decreased, as ESDD increased. For the NaCl samples with the same ESDD, the spectral intensity of Na was higher in the NaCl sample with a particle size of 60.914 μm than that in the NaCl sample with a particle size of 240.764 μm, exhibiting a trend similar to that in Figure 8.

**Figure 10.** Spectral intensities of Na at 589 nm.

**Figure 11.** Changes in the spectral intensities of Na in NaCl samples differing in particle size with NaCl concentration.

Density is one of the variable properties of pollutions. Pollutions on the surfaces of insulators in different operating environments vary significantly in density. Thus, it is necessary to examine the effects of pollution density on LIBS spectral signals. Four identical artificial pollution samples, each consisting of kaolin clay (2 g) and NaCl (1%), were prepared. The four samples were compressed using a compression machine under compressive loads of 6, 9, 12 and 15 t. The compressed samples were subsequently subjected to LIBS testing to determine the relationship between compressive load and average relative spectral intensity (Figure 12). As demonstrated in Figure 12, as the density increased, the relative excited spectral intensities of the samples increased. This phenomenon can be explained by excited plasma plume dynamics. When the laser energy acts on the surface of a sample, the denser the surface of the sample is, the greater the impact of the laser pulse reverse shock wave is. Various types of particles jet from the target surface opposite the direction of the laser. The increase in the reverse jet velocity and intensity of various types of particles strengthens the collision ionization during the rapid expansion of the plasma, thereby, improving the atomic emission intensity.

**Figure 12.** Effects of density on spectral intensity.

#### **3. Experiments**

A composite insulator chain (manufactured by Dongguan Gaoneng Industry Co., Ltd. in Dongguan, China) was collected from the N63 jumper of the 220-kV Dongguan–Kuihu line A. As shown in Figure 13, an insulator was cut from the centre of the insulator chain along the external surface of the core of the chain. A small piece (1 cm × 1 cm) was cut from a relatively dark-colored area of the insulator and subjected to scanning electron microscopy (SEM) and energy dispersive X-ray spectroscopy (EDS) analysis on a Zeiss Supra 55 SEM (manufactured by Carl Zeiss Co., Ltd. in Oberkochen, Germany) equipped with an Oxford X-Max 20 EDS detector (manufactured by Oxford Instruments Co., Ltd. in Oxford, Britain) to determine the content and distribution of the pollutions on the surface. A Leica EMACE 200 fully automatic low-vacuum coating system was used to coat the sample with Pt to improve its surface conductivity.

**Figure 13.** Schematic diagram of the insulators used in the experiment.

A LIBS system assembled by our research group was used in the experiment. This LIBS system consists of a Nimma-900 laser system (wavelength: 1,064 nm, pulse width: 10ns, output frequency: 1 Hz, and output energy: 110 mJ), focal spot diameter of approximately 80 μm, laser energy density of 2.1883 <sup>×</sup> 10<sup>11</sup> Watts/cm2, an Avantens spectrometer (available wavelength range: 200–650 nm) and a DG645 delay controller. The delay controller controls the interval between the output of the laser system and the acquisition of the spectrometer to effectively obtain an atomic emission spectrum evolved from a continuous background emission spectrum generated after plasma excitation under the action of the laser. The delay time was set to 3 μs in the experiment. The horizontal laser beam emitted by the laser system was reflected by a 45◦ mirror onto the vertical plane and focused by a convex lens onto the surface of the sample. The lens-to-sample distance was adjusted to position the sampling spot at the focal point of the convex lens. The spectral data acquired by the spectrometer were exported using the software Avasoft 8.8 (developed by Avantes Co., Ltd. in Apeldoorn, the Netherlands) and were subsequently processed.

In the experiment, kaolin clay was mixed with NaCl of different particle sizes at a 1:1 mass ratio. The shape of mixture is a circle with a diameter of 8 nm. Each mixture was compressed using a compression machine under compressive loads of 9t. and subsequently subjected to LIBS testing. Thus, the spectra of the Na in NaCl of different particle sizes were obtained. The NaCl particle size was determined using a laser particle size analyzer. In addition, the spectral intensity of NaCl obtained by LIBS was normalized to improve the analytical accuracy.

Spectrographic-grade NaCl samples (manufactured by Aladdin Industrial Co., Ltd. in Shanghai, China) were used in the experiment. Additionally, after sieving through <60, 60–100, 100–200 and 200–300 mesh stainless steel sieves, corresponding to particle sizes of >250, 150–250, 75–150 and 200–300 μm, respectively, NaCl samples of four different particle sizes were obtained (50 g of each type).

#### **4. Conclusions**

In this study, the microregional characteristics and element distributions of natural pollutions were analyzed with LIBS. The conclusions derived from this study are summarized as follows:


In this study, the laser energy was set to 80 mJ, corresponding to a laser energy ablation density of 3.814 <sup>×</sup> 1010 Watts/cm2.

(4) The effects of the pollution properties (particle size and density) on the spectral signals were analyzed. A decrease in the particle size and an increase in the density of the sample both, improved the relative spectral intensities of the elements tested.

**Author Contributions:** Conceptualization and formal analysis, X.W. (Xinwei Wang) and S.L.; investigation, X.Q.; resources, X.W. (Xinwei Wang) and T.W.; data curation, S.L. and T.W.; writing—original draft preparation, X.Q.; writing—review and editing, X.W. (Xilin Wang) and Z.J.; supervision, X.W. (Xilin Wang) and Z.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by National Natural Science Foundation of China (51607101), Science and technology projects of Shanxi Electric Power Research Institute (SGSXDK00SPJS1900162), and the Guangzhou Science and Technology Plan (201707020044).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Sample Availability:** The composite insulator chain and samples of the NaCl in different particle size are available from the authors.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Development of a Direct Competitive ELISA Kit for Detecting Deoxynivalenol Contamination in Wheat**

**Li Han 1,2,**†**, Yue-Tao Li 2,**†**, Jin-Qing Jiang 2, Ren-Feng Li 2, Guo-Ying Fan 2, Jun-Mei Lv 2, Ye Zhou 2, Wen-Ju Zhang 1,\* and Zi-Liang Wang 2,\***


#### Academic Editor: Clinio Locatelli

Received: 4 November 2019; Accepted: 18 December 2019; Published: 22 December 2019

**Abstract:** This study was conducted to develop a self-assembled direct competitive enzyme-linked immunosorbent assay (dcELISA) kit for the detection of deoxynivalenol (DON) in food and feed grains. Based on the preparation of anti-DON monoclonal antibodies, we established a standard curve with dcELISA and optimized the detection conditions. The performance of the kit was evaluated by comparison with high-performance liquid chromatography (HPLC). The minimum detection limit of DON with the kit was 0.62 ng/mL, the linear range was from 1.0 to 113.24 ng/mL and the half-maximal inhibition concentration (IC50) was 6.61 ng/mL in the working buffer; there was a limit of detection (LOD) of 62 ng/g, and the detection range was from 100 to 11324 ng/g in authentic agricultural samples. We examined four samples of wheat bran, wheat flour, corn flour and corn for DON recovery. The average recovery was in the range of 77.1% to 107.0%, and the relative standard deviation (RSD) ranged from 4.2% to 11.9%. In addition, the kit has the advantages of high specificity, good stability, a long effective life and negligible sample matrix interference. Finally, wheat samples from farms in the six provinces of Henan, Anhui, Hebei, Shandong, Jiangsu and Gansu in China were analyzed by the kit. A total of 30 samples were randomly checked (five samples in each province), and the results were in good agreement with the standardized HPLC method. These tests showed that the dcELISA kit had good performance and met relevant technical requirements, and it had the characteristics of accuracy, reliability, convenience and high-throughput screening for DON detection. Therefore, the developed kit is suitable for rapid screening of DON in marketed products.

**Keywords:** deoxynivalenol; dcELISA kit; performance measurement; development

#### **1. Introduction**

Deoxynivalenol (DON), also known as vomitoxin, is a highly toxic secondary metabolite produced by *Fusarium graminearum* and *Fusarium culmorum*; DON belongs to the B-group of trichothecenes and widely exists in various agricultural products, food, and animal feed, especially in wheat, maize, and other cereal crops [1–4]. DON readily acts as an animal antifeedant and shows immunotoxicity, organ toxicity, inhibition of protein synthesis, and teratogenicity. These symptoms are closely related to immune suppression, Keshan disease, oesophageal cancer and other diseases [5–7]. Moreover, DON is heat-stable, and general cooking and processing cannot destroy its toxicity. Young et al. [8] found that grain processed into pet food still contained DON. Therefore, DON pollution poses a great threat to human and livestock health and has attracted the attention of countries around the world [9]. At present, at least 100 countries have mandatory limits on DON levels in food and feed. In view of its serious toxic effects, in the preliminary draft of the DON maximum levels (MLs), the Codex Alimentarius Food (FAO) Committee recommended the following limits: 2 mg/kg in unprocessed cereals, 1 mg/kg in semi-processed products using wheat, corn and barley as raw materials, and 0.5 mg/kg in cereals for infants and young children [10,11]. In China, the ML of DON in maize, wheat, and their products is regulated at 1 mg/kg [12,13]. The molecular formula of DON is C15H20O6, and its molecular weight is 296.32 [14]. Its structure is shown in Figure 1:

**Figure 1.** Molecular structure of deoxynivalenol (DON).

At present, the main physical and chemical methods for detecting DON contamination in food and feed are thin-layer chromatography (TLC), high-performance liquid chromatography (HPLC), gas chromatography (GC), mass spectrometry (MS), gas chromatography-mass spectrometry (GC-MS), high-performance liquid chromatography–mass spectrometry (HPLC–MS), high-performance liquid chromatography–tandem mass spectrometry (HPLC–MS/MS), and others [15–17]. These methods have high precision and sensitivity, but the sample pretreatment is rigorous, the instruments are expensive, the detection range is small, and, as the analysts often need special training, the cost is high. These methods are only suitable for large enterprises, scientific research institutes, or testing institutions that require high detection sensitivity, and they are not suitable for the demand of DON pollution detection in the feed industry. Therefore, increasing attention has been paid to the simple, rapid, sensitive, low-cost enzyme-linked immunosorbent assay (ELISA), which is suitable for large-scale sample screening. For example, the traditional immunosorbent assay (ELISA) [18], chemiluminescence enzyme immunoassay (CLEIA) [19], fluorescence polarization immunoassay (FPIA) [20], time-resolved fluorescence immunoassay (TRFIA) [21], colloidal gold immunochromatography (GICA) [22,23], surface plasmon resonance (SPR) immunoassay [24], silver-stained GICA [25], nanobody-based ELISA [26], and immunosensor, among others, can be used to detect DON. Therefore, due to the prevalence of DON contamination and the large number of samples that need to be analyzed, ELISA kits have been considered a suitable detection tool, and their development and application has grown rapidly in recent years because they do not need special instruments and equipment, are suitable for the field and are suitable for high-throughput screening.

The purpose of this experiment is to assemble and optimize a new DON dcELISA kit. The performance of the ELISA kit was tested, and its accuracy was verified by HPLC, which laid a foundation for the development of ELISA kits with high sensitivity, specificity, and good quantification suitable for screening a large number of DON-contaminated samples.

#### **2. Materials and Methods**

#### *2.1. Reagents and Materials*

The standards of DON, 3-Ac-DON, 15-Ac-DON, Nivalenol (NIV), Fusarenon-X, T-2 toxin, Zearalenone (ZEN), and Aflatoxin B1 (AFB1) were purchased from Sigma-Aldrich Co., Ltd. (Augsburg, Germany). Bovine serum albumin (BSA), chicken ovalbumin (OVA), *N*,*N* -carbonyldiimidazole (CDI), anhydrous tetrahydrofuran (THF), *N*,*N*-dimethylformamide (DMF), horseradish peroxidase (HRP), 1-ethyl-3-(3-dimethylamino)propyl) carbodiimide hydrochloride (EDC), Freund's complete adjuvant (FCA), and Freund's incomplete adjuvant (FIA) were provided by Pierce. PEG-1500 (polyethylene glycol) was purchased from Roche. GaMIgG was purchased from Huamei Biotechnology Company (Shanghai, China). In addition, 96-well microtiter plates as well as 24-well and 96-well cell culture plates were purchased from Iwaki Co., Ltd. (Dalian, China); 3,3,5,5-Tetramethylbenzidine (TMB), phenacetin and urea peroxide were purchased from Sigma. Foetal bovine serum (FBS) was purchased from Gibco. Female Balb/c mice (6 to 8 weeks old) were provided by Beijing SPF Biotech Co., Ltd. (Beijing, China) and were raised under strict control in our laboratory animal house.

Phosphate-buffered saline (PBS), carbonate-buffered saline (CBS), washing buffer (PBST, PBS containing 0.05% Tween-20), blocking buffer (SPBST, PBST containing 5% goat serum), color substrate solution (TMB), stopping solution (2 M H2SO4), Glucose sodium chloride potassium chloride solution (GNK), complete medium, Hypoxantin Aminopterin and Thymidin (HAT) medium, Hypoxantin and Thymidin (HT) medium, were all made in-house in our laboratory.

A Galaxy S-type CO2 cell incubator was purchased from Biotech. A Multiskan MK3 microplate reader was purchased from Thermo (Waltham, Ma, USA) and used for 450 nm absorbance measurements. An inverted MIC 00949 microscope was purchased from Nikon Corporation. A DK-8D water bath was provided by Yiheng Instrument Co., Ltd. (Shanghai, China). A BS124S electronic balance was purchased from the German Sartorius Group. Purified water was prepared using a Milli-Q purification system (Millipore Corporation, Bedford, MA, USA). An A11 basic analytical mill was provided by IKA (Staufen, Germany). A Legend Micro 17 centrifuge was provided by Thermo (Waltham, MA, USA). Glass microfiber filter paper was purchased from Whatman (Maidstone, UK). The reliability of the ELISA kit was confirmed using an Agilent 1260 HPLC equipped with a diode array detector (DAD) (Agilent Technologies, Wilmington, DC, USA).

#### *2.2. Preparation of the Antigen and Anti-DON Monoclonal Antibody (mAb)*

According to the molecular structure of DON, the artificial antigen DON-BSA was synthesized by the carbonyl diimidazole (CDI) procedure outlined in a previously published method by Maragos et al. [14], with slight modifications. The synthesis of coated DON-OVA was improved by referring to the method of Li et al. [27]. DON was derivatized by maleic anhydride, and then the hapten was coupled with OVA to coat the original DON-OVA by implementation of the carbodiimide (EDC) procedure. Preparation of anti-DON mAb was achieved with classical hybridoma technology [28]. After obtaining DON mAb hybridoma cell lines, this experiment adopted an in vivo induced ascites method [29] to mass produce DON mAb, which was then purified from ascites by an octanoic acid/ammonium sulfate precipitation method [30]. The DON mAb was then stored at −20 ◦C until the dcELISA kit was assembled.

#### *2.3. Development of the DON dcELISA Kit*

We prepared the enzyme-labeled hapten (horseradish peroxidise-DON) and determined its working concentration as follows.

The enzyme-labeled hapten (HRP-DON) was prepared by a carbonyl diimidazole (CDI) method. DON standard (5 mg) was dissolved in 1 mL THF, 60 mg CDI was added, and the reaction proceeded for 4 h in a dry environment at 70 ◦C. The solvent of the reaction products was evaporated, and 500 μL DMF was added to the remaining products and completely dissolved. Then, 2 mg of HRP was added dropwise (the HRP was dissolved in 2 mL 0.01 mol/L pH 7.4 PBS solution) and stirred for 24 h at 4 ◦C in the dark. The reaction products were dialyzed in PBS for 72 h, the fluids were replaced 9 times during dialysis, and the dialysate, which was the enzyme-labeled hapten HRP-DON, was collected.

It is well known that the working concentration of the coated antigen and antibody is the key to determine the sensitivity of an ELISA kit. To determine the optimum dilution of RaMIgG, anti-DON mAb, and HRP-DON, chessboard titration tests were carried out. HRP-DON was added to 50% glycerol and stored at −20 ◦C.

#### *2.4. Components of the ELISA Kit*

The optimum conditions of the kit were very important for improving detection technology. The components and parameters of the kit are shown in Table 1 [31]:

**Table 1.** Components and parameters of the direct competitive enzyme-linked immunosorbent assay (dcELISA) kit.


#### *2.5. Establishment of the Kit Standard Curve*

The standard curve was established by dcELISA. The inhibition rate B/B0 of a series of concentrations of DON standards against DON mAb was taken as the ordinate, and the logarithmic value of a series of concentrations of DON standards was taken as the abscissa. The standard curve was analyzed and fitted using Origin Program 7.0 software (OriginLab Co., Northampton, MA, USA), and the linear regression was established. The theoretical detection limit and linear detection range of the kit were calculated by the regression equation.

#### *2.6. Pretreatment of Samples*

The wheat samples came from farms in six Chinese provinces: Henan, Anhui, Hebei, Shandong, Jiangsu and Gansu. A total of 30 samples were randomly checked (5 samples from each province). After the samples were ground, 5 g of each sample was accurately weighed (accurate to 0.01 g) and placed in a bottle. Distilled water (25 mL) was added, and extracted by sonication for 10 min. The mixture was evenly mixed for a few minutes. The supernatant was centrifuged at 8000 rpm/min for 5 min. Finally, 500 μL of the supernatant was added to 500 μL of the sample diluent, which is the extract solution of the sample to be tested. In addition, the pH of the sample extract was adjusted from 6 to 8. If needed, samples were diluted with the working buffer before being analyzed with the kit.

#### *2.7. Operating Procedure of the Kit*


#### *2.8. Characteristics of the DON dcELISA Kit*

#### 2.8.1. Sensitivity Determination

According to the method of Hayashi et al. [32], the sensitivity of competitive ELISA is B/B0% =83.3%; the sensitivity of the kit was calculated according to the standard curve regression equation, and the detection limit was also determined.

#### 2.8.2. Accuracy and Precision Determination

In this study, the accuracy and precision of the kit were determined with recovery experiments and expressed as recovery (%) and relative standard deviation (RSD%), respectively. The wheat bran, wheat flour, maize flour, and maize were first treated with 1% Na2CO3 for detoxification [33]. Then, 5 g of each sample was spiked with DON at 200, 500 and 1000 ng/g and stirred for 2 h at room temperature (RT). Next, the spiked samples were added to 10 mL of working buffer containing 20% methanol, and extracted by sonication for 10 min. The supernatant was centrifuged at 8,000 rpm/min for 5 min. Finally, 500 μL of the supernatant was added to 500 μL of the working buffer, which is the extract solution of the sample to be tested. Then, each sample was tested three times, and the recovery (%) and RSD% were calculated:

$$\text{Recovery } (\%) = \text{the measured value/the actual added value} \times 100\% \tag{1}$$

$$\text{RSD} \left( \% \right) = \text{SD} \left( \text{standard deviation} \right) / \text{\AA} \left( \text{mean value} \right) \times 100\% \tag{2}$$

#### 2.8.3. Specificity Determination

The specificity of the cross-reactions between the kit and other mycotoxins was evaluated, and the formula of cross-reaction rate (CR%) is [34]:

$$\text{CR (\%)} = \text{[IC}\_{50} \text{ (DON)} \text{/IC}\_{50} \text{ (Structural Analysis)}] \times 100\% \tag{3}$$

#### 2.8.4. Stability Determination

The stability of the kit was evaluated by the changes in B0 (the value of absorbance without the DON standard) and B/B0 (%) (the ratio value of absorbance with 5 ng/mL DON and without the DON standard) during storage (2 to 8 ◦C).

#### 2.8.5. Matrix Effect Determination

To analyze the effect of the sample matrix on the sensitivity of the kit, the DON standard solution was dissolved in four samples of wheat bran, wheat flour, corn flour, and corn. These samples were then diluted with sample diluent, and the curve was generated according to the operation of the kit.

#### *2.9. Confirmation of the DON dcELISA Kit with HPLC*

The wheat samples from farms in the six provinces of Henan, Anhui, Hebei, Shandong, Jiangsu and Gansu in China were tested using the assembled DON dcELISA kit and HPLC. A total of 30 samples were randomly checked (5 samples from each province), and the correlation between the kit and HPLC was evaluated by comparing the results of detection [35]. Sample extraction and HPLC analysis were performed according to the method of the national standard of China GB5009.111-2016 [36], with slight modifications. After the samples were ground, 5 g of each sample was accurately weighed (accurate to 0.01 g) and placed in a clean and capped wide-mouth bottle. Twenty-five milliliters of acetonitrile-H2O (20:80, *v*/*v*) and 2 g polyethylene glycol were added. The bottle was capped and extracted by sonication for 30 min. The mixture was evenly mixed for a few minutes. The samples were centrifuged at 6000 rpm/min for 10 min. Finally, the supernatants were filtered through glass microfiber filters to clarify the extract solution of the sample to be tested. Then, the supernatants were purified through DON immunoaffinity columns. The extracted phases were collected and analyzed by HPLC. The HPLC analysis was performed using an Agilent 1260 HPLC equipped with a diode array detector (DAD). Separation was performed on a C18 liquid chromatographic column (150 mm × 4.6 mm × 5 μm) or equivalent, the mobile phase was methanol:water (20:80, *v*/*v*), the flow rate was 0.8 mL/min, the column temperature was 35 ◦C, the injection volume was 50 μL, and the detection wavelength was 218 nm.

#### **3. Results**

#### *3.1. Development of the DON dcELISA Kit*

For the determination of the working concentrations of RaMIgG, anti-DON mAb and HRP-DON using the chessboard titration tests, ELISA microplates were coated with 10 ng/mL RaMIgG. The working concentrations of the anti-DON mAb and HRP-DON were determined as 1:6400 (1.56 ng/mL) and 1:800 (28.5 ng/mL), respectively, when the value of B0 reached 1.0.

The key parameters were studied to guarantee the ideal sensitivity and performance of the kit for detecting DON. Under the criteria of a higher value of B0/half-maximal inhibition concentration (IC50) and lower value of IC50, the working buffer, which could greatly affect the sensitivity of the kit, was adjusted. Finally, 5% methanol, 0.5 mol/L Na+, and pH 7.4 in the working buffer were selected as the optimal working buffer for the kit (Table 2).



#### *3.2. Generating and Fitting the Standard Curve of the Kit*

The standard curve of the kit is shown in Figure 2. By analyzing the curve, the regression equation <sup>y</sup> <sup>=</sup> <sup>−</sup>32.433x <sup>+</sup> 76.608, correlation coefficient R2 <sup>=</sup> 0.972, and IC50 <sup>=</sup> 6.61 ng/mL was obtained, and the detection range (IC10 to IC80) was 1.0 to 113.24 ng/g.

**Figure 2.** Calibration curve of the dcELISA kit.

#### *3.3. Performance Measurements of the Kit*

#### 3.3.1. Sensitivity Determination

When B/B0 = 83.3%, the corresponding DON concentration was 0.62 ng/g, indicating a sensitivity of 0.62 ng/g, which was obtained by substituting the B/B0 value into the standard curve regression

equation. However, considering the need for positive detection and the error of user operation, the detection limit of the competitive ELISA kit was determined to be 1.0 ng/g.

#### 3.3.2. Accuracy and Precision Measurement

Table 3 shows the four feed samples of wheat bran, wheat flour, corn flour, and corn with the recoveries. The average recovery was in the range of 77.1% to 107.0%, and the RSD ranged from 4.2% to 11.9%. The dcELISA kit meets the requirements of national accuracy and precision, indicating that the kit can be used for the detection of actual samples.


**Table 3.** Recoveries of DON in different samples by the dcELISA kit (*n* = 3).

#### 3.3.3. Specificity Determination

Table 4 shows that the cross-reactions between the kit and other mycotoxins were negligible. The cross-reaction rate with 3-Ac-DON was 4.7% and that with other mycotoxins was less than 0.2%, indicating that the kit has high specificity.


**Table 4.** Cross-reactivity of the DON dcELISA kit with other related mycotoxins.

#### 3.3.4. Stability Determination

As shown in Figure 3, the values of B0 (the value of absorbance without DON standard) and B/B0 (%) (the ratio value of absorbance with 5 ng/mL DON and without DON standard) showed acceptable decreases during storage. The results showed that the kit had good stability and that its effective life was at least 12 months.

**Figure 3.** Stability of the dcELISA kit.

#### 3.3.5. Matrix Effect Determination

As shown in Figure 4, the curves of the spiked samples of wheat bran, wheat flour, corn flour and corn were close to the DON standard curve by dilution of the extract solution multiple times, and their IC50 values were 8.81, 7.59, 6.22 and 5.7 ng/mL, respectively, indicating that the matrix interference was negligible. Therefore, the kit is functional for different substrates and can be used for detecting subsequent samples.

**Figure 4.** Effect of different samples' matrixes on the dcELISA kit.

#### *3.4. Confirmation of the DON dcELISA Kit with HPLC*

Table 5 shows that a total of 30 wheat samples from different provinces in China were tested using the assembled DON dcELISA kit and HPLC. The average value of detection with HPLC was in the range of 560.4 to 1049.1 ng/g, and the RSD ranged from 12.4% to 43.4% (the results of HPLC were corrected by a recovery of 85.7%). The average value of detection with the kit was in the range of 580.5 to 1020.3 ng/g, and the RSD ranged from 13% to 43.8%. The results showed that the test results of the kit were generally higher than those of HPLC. However, the test results of the kit in its linear range were in good agreement with those of HPLC.


**Table 5.** Comparison of screening results of 30 wheat samples detected by two different methods.

Thirty samples of wheat were detected with the kit, 22 of which were found to contain DON, with a concentration range of 254.7to 1258.4 ng/g. Four of the 30 samples were false suspect, with a false suspect rate of 13.3% (Table 6).

**Table 6.** Test results of wheat samples from different provinces by the kit and HPLC.


+, positive; -, negative; +, false negative; **-**, false suspect.

#### **4. Discussions**

#### *4.1. Pretreatment of Biotoxin Samples*

The pretreatment of samples increases the accuracy of HPLC and ELISA analyses. The samples were extracted and detected for DON; the pretreatment of samples for detection by ELISA was relatively simple, and direct filtration after extraction was sufficient for detection, while the pretreatment of samples for detection with HPLC required an immunoaffinity column. Yang et al. reported [37] that, the recovery rate of DON in ELISA was higher than 75%, and the RSD was 4.7% to 10.6% after the sample was filtered directly, while after passing through the immunoaffinity column, the recoveries

of DON in HPLC and ELISA were the same when the spiked concentration of the standard was higher. It is concluded that the DON kit simplifies the process of sample pretreatment and purification. The results are accurate and reliable, and the detection steps are simple. It is very suitable for the rapid detection of a large number of samples. Moreover, ELISA detection technology has the advantages of limited interference, strong specificity, and short enzymatic reaction times, which shortens the whole detection time.

#### *4.2. Determination of dcELISA Kit Performance*

In this study, the dcELISA kit was assembled with an in-house-developed homemade high-affinity anti-DON monoclonal antibody. The kit performance metrics included sensitivity, accuracy, precision, specificity, stability and matrix effect, among others. Sensitivity determination can be calculated according to the method of Hayashi et al. [32]. The sensitivity of competitive ELISA is B/B0 = 83.3%, which can also be calculated by the formula of limit of detection LOD (%) = [(X − 2SD)/X] × 100%. The method of B/B0 = 83.3% was adopted in this experiment. The sensitivity was determined as 0.62 ng/mL, the detection limit was 1.0 ng/mL, and the detection range (IC10 to IC80) was 1.0 to 113.24 ng/mL in the working buffer. According to the procedures of authentic sample pretreatment and extraction, the DON levels of samples were equivalent to a 100-fold dilution and the matrix effects were negligible. Thus, for the analysis of wheat samples, with a sensitivity of 62 ng/g, an LOD of 100 ng/g, and a detection range from 100 to 11,324 ng/g in authentic agricultural samples, the cross-reaction rate with 3-Ac-DON and 15-Ac-DON was 4.7%, less than 0.2%, respectively. The DON ELISA method established by the Ministry of Health in China has a detection limit of 5 ng/mL, and the detection range was 5 to 1000 ng/mL. It had been approved as the national recommended standard detection method of China. Therefore, the DON dcELISA kit assembled in our laboratory meets the domestic detection range and sensitivity standard requirements of DON analysis in food and feed. Compared with the commercial kits, the sensitivity and specificity is higher (3 ng/mL in the working buffer), the DON levels of samples were equivalent to a 100-fold dilution with an LOD of 300 ng/g in authentic agricultural samples, and the cross-reaction rate with 3-Ac-DON and 15-Ac-DON was less than 70%, less than 1%, respectively. Compared with the kit that was developed by Li et al. [12], it has higher sensitivity and specificity (4.9 ng/mL in the working buffer), the DON levels of samples were equivalent to a 40-fold dilution with an LOD of 200 ng/g in authentic agricultural samples, and the cross-reaction rate with 3-Ac-DON and 15-Ac-DON was 5.7%, less than 0.5%, respectively. Accuracy and precision are measured by spiked sample recovery (%) and relative standard deviation (RSD%). Generally, the recovery rate is between 70% and 140%. The average recovery was in the range of 77.1% to 107.0%, and the RSD was 4.2% to 11.9% in this experiment, which meets the national accuracy and precision test requirements, indicating that the kit could be used for the detection of actual samples. Therefore, by evaluating the recoveries and determining the DON content in wheat samples, it is proved that the developed dcELISA kit is accurate, reliable, and simple, and that it requires less instrumentation, and involves simple experimental steps for detecting DON content in food and feed. Compared with commercial kits, it is a more advanced detection method in China and abroad, providing a highly sensitive, economical and safe DON detection method.

#### *4.3. Comparison of the Results of the dcELISA Kit and HPLC*

The kit test results were generally higher than those of HPLC. The wheat samples from the farms in the six provinces of Henan, Anhui, Hebei, Shandong, Jiangsu and Gansu in China were analyzed for DON content using both the kit and HPLC. A total of 30 samples were randomly checked (five samples from each province). The average value of detection with HPLC was in the range of 560.4 to 1049.1 ng/g, and the RSD ranged from 12.4% to 43.4%. The average value of detection with the kit was in the range of 580.5 to 1020.3 ng/g, and the RSD ranged from 13% to 43.8%. Therefore, the results showed that the test results of the kit were generally higher than those of the HPLC. However, the test results of the kit in its linear range were in good agreement with those of HPLC. Therefore, the kit can be used for the determination of DON in food and feed. Antibodies are the basis of the ELISA kit detection method, which may lead to false positive or false negative results, while HPLC is commonly used as an accurate verification method. There were two main reasons why the kit test results were generally too high. First, high or low pH of the sample solution will affect the test results. Therefore, it is necessary to adjust the pH of the sample solution before detection. Some studies have found that when the pH of a sample extract is lower than 5, the structure of the enzyme in an enzyme-labeled antigen changes irreversibly, and most of its activity is lost, resulting in a reduced color reaction, which leads to false-positive results. The pH of the extracted solution was from 6 to 8 when the samples were purified in this experiment, which met the detection requirements of the kit. Therefore, the pH of the samples did not need to be adjusted when samples were detected. Second, the loss of sample in the pretreatment process of HPLC leads to low detection results (the results of HPLC in this experiment were corrected by a recovery of 85.7%).

#### **5. Conclusions**

In this experiment, a dcELISA kit method was established by using an anti-DON monoclonal antibody developed in our laboratory. The working concentrations of RaMIgG, anti-DON mAb and HRP-DON were optimized, and the performance of the developed kit was tested. Finally, a comparison of the results of the kit with those of HPLC shows that the developed kit has the same detection ability as HPLC. Therefore, the kit can be widely used for DON detection in food and feed.

**Author Contributions:** W.-J.Z. and Z.-L.W. designed the research and interpreted the results. L.H. and Y.-T.L. conducted the experiments and drafted the manuscript. Y.Z. and J.-M.L. collected and processed data. J.-Q.J., R.-F.L. and G.-Y.F. revised the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was financially supported by the Modern Agricultural Science and Technology Tackling and Achievement Conversion Project of the Eighth Normal University [2018NY05]; the Twelfth Five-Year Plan of National Science and Technology Support Projects, "Research and Demonstration of Rapid Detection Technology for Hormone Drugs" [2014BAD13B05-01]; the Key Technological Research Projects of Henan Province in 2018 (Agriculture) [182102110222]; and the Program for Innovative Research Team (in Science and Technology) in University of Henan Province (20IRTSTHN025).

**Conflicts of Interest:** There are no conflict to declare.

#### **References**


**Sample Availability:** Samples of the compounds are not available from the authors.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
